Wavelet transform-based power grid fault positioning method and device

Abstract

The embodiment of the invention discloses a power grid fault positioning method and device based on wavelet transform, and the method comprises the steps: collecting voltage, current and power flow signals of a power grid, and generating a power grid dynamic signal data set containing time sequence features; performing multi-scale decomposition on the power grid dynamic signal data set by using wavelet transform based on signal local complexity optimization to generate initial signal feature components under each scale; fractal geometric analysis is carried out on the initial signal feature component, and a corresponding Hurst index and a fractal dimension are calculated; calculating a Lyapunov index of the initial signal feature component, and dynamically adjusting scale parameters of fractal analysis according to the Lyapunov index; constructing an enhanced feature set representing fault features according to the optimized Hurst index and the calculation result of the fractal dimension; and based on the enhanced feature set, constructing a dynamic topology model of the power grid fault, and generating a fault positioning result. According to the invention, the performance of power grid fault detection and positioning is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of power system fault diagnosis, in particular to a power grid fault positioning method and device based on wavelet transform. BACKGROUND

[0002] Power grid fault positioning is a key technology to ensure the stability and reliability of power systems. Traditional methods usually rely on time-domain or frequency-domain analysis, which is difficult to effectively capture the complex, nonlinear and multi-scale characteristics of modern power grid signals. For example, methods based on Fourier transform or simple statistical analysis often lack accuracy when dealing with dynamic and chaotic signals. Wavelet transform can capture local anomalies in signals through multi-scale decomposition, but its feature extraction is easily affected by noise interference or improper scale selection, limiting the robustness of fault features.

[0003] Therefore, there is an urgent need for a method to improve the performance of power grid fault detection and positioning by modifying the feature extraction of wavelet transform and constructing a dynamic topology model. SUMMARY

[0004] The present application provides a power grid fault positioning method and device based on wavelet transform, which improves the performance of power grid fault detection and positioning.

[0005] The present application provides the following solutions:

[0006] According to a first aspect, a power grid fault positioning method based on wavelet transform is provided, the method comprising: collecting voltage, current and power flow signals of the power grid to generate a power grid dynamic signal dataset containing time series features; using wavelet transform based on signal local complexity optimization to perform multi-scale decomposition on the power grid dynamic signal dataset to generate initial signal feature components at each scale; performing fractal geometry analysis on the initial signal feature components to calculate corresponding Hurst exponents and fractal dimensions; calculating the Lyapunov exponents of the initial signal feature components and dynamically adjusting the scale parameters of fractal analysis according to the Lyapunov exponents to obtain the calculation results of the optimized Hurst exponents and fractal dimensions; constructing an enhanced feature set representing fault features according to the calculation results of the optimized Hurst exponents and fractal dimensions; based on the enhanced feature set, constructing a dynamic topology model of power grid faults, and identifying the node or line location of fault occurrence according to the model to generate a fault positioning result

[0007] According to an implementable manner in the embodiments of the present application, the multi-scale decomposition of the power grid dynamic signal dataset by using wavelet transform based on local complexity optimization of signals to generate initial signal feature components at each scale includes: calculating the local Shannon entropy of the power grid dynamic signal dataset to determine the local complexity of the signal; dynamically selecting a wavelet basis function and a decomposition layer number according to the local complexity, and performing wavelet transform to generate multi-scale initial signal feature components.

[0008] According to an implementable manner in the embodiments of the present application, the fractal geometry analysis of the initial signal feature components to calculate the corresponding Hurst index and fractal dimension includes: using a box-counting-based fractal analysis method to calculate the fractal dimension of the initial signal feature components at each scale; and calculating the Hurst index by range analysis to represent the long-range correlation of the signal.

[0009] According to an implementable manner in the embodiments of the present application, calculating the Lyapunov exponent of the initial signal feature components and dynamically adjusting the scale parameter of the fractal analysis according to the Lyapunov exponent includes: calculating the Lyapunov exponent of the initial signal feature components by Wolf algorithm to quantify the chaotic characteristics of the signal; and dynamically adjusting the window scale of the fractal analysis according to the size of the Lyapunov exponent, so that the window size is positively correlated with the chaotic degree of the signal.

[0010] According to an implementable manner in the embodiments of the present application, constructing an enhanced feature set representing fault characteristics according to the calculation results of the optimized Hurst index and fractal dimension includes: performing weighted fusion on the calculation results of the optimized Hurst index and fractal dimension by using a nonlinear mapping based on a Gaussian kernel to obtain a fusion feature; and using mutual information entropy analysis to screen a feature subset with high information quantity in the fusion feature to form the enhanced feature set.

[0011] According to an implementable manner in the embodiments of the present application, constructing a dynamic topology model of the power grid fault based on the enhanced feature set and identifying the node or line location where the fault occurs according to the model includes: constructing a weighted directed graph model of the power grid based on the enhanced feature set, wherein the nodes represent the power grid equipment and the edges represent the line connections; using a graph attention network to calculate the feature correlation of each node and edge in the weighted directed graph by a multi-head attention mechanism to generate dynamically allocated attention weights; calculating the fault probability of each node and line by iteratively optimizing the weight parameters of the graph attention network; and determining the node or line location where the fault occurs according to a preset probability threshold to output the fault positioning result.

[0012] According to an implementable manner in the embodiments of the present application, the output of the fault locating result further includes: constructing a statistical model of fault probability distribution based on historical operation data of the power grid dynamic signal data set; dynamically adjusting the probability threshold based on the statistical model and in combination with the current power grid operation state; and reselecting the fault node or line according to the adjusted probability threshold to generate an optimized fault locating result.

[0013] According to a third aspect, a power grid fault locating apparatus based on wavelet transform is provided, which comprises: a power grid data acquisition unit configured to acquire voltage, current and power flow signals of a power grid to generate a power grid dynamic signal data set containing time series features; a feature component generation unit configured to perform multi-scale decomposition on the power grid dynamic signal data set by using wavelet transform based on signal local complexity optimization to generate initial signal feature components at each scale; a fractal geometry analysis unit configured to perform fractal geometry analysis on the initial signal feature components to calculate corresponding Hurst indexes and fractal dimensions; a fractal result optimization unit configured to calculate Lyapunov indexes of the initial signal feature components and dynamically adjust scale parameters of fractal analysis according to the Lyapunov indexes to obtain calculation results of optimized Hurst indexes and fractal dimensions; an enhanced feature set construction unit configured to construct an enhanced feature set representing fault features according to the calculation results of the optimized Hurst indexes and fractal dimensions; and a locating result generation unit configured to construct a dynamic topology model of power grid faults based on the enhanced feature set and identify node or line positions where faults occur according to the model to generate a fault locating result.

[0014] According to a third aspect, a computer readable storage medium is provided, which stores a computer program, and the program is executed by a processor to implement the steps of the method according to any one of the first aspect.

[0015] According to a fourth aspect, an electronic device is provided, which comprises: one or more processors; and a memory associated with the one or more processors, the memory being configured to store program instructions, and the program instructions being configured to perform the steps of the method according to any one of the first aspect when being read and executed by the one or more processors.

[0016] According to the specific embodiments provided by the present application, the following technical effects are disclosed:

[0017] The application generates a dynamic signal data set by collecting power grid voltage, current and power flow signals, performs multi-scale decomposition combined with optimized wavelet transform to extract initial signal characteristic components. The Hurst index and fractal dimension are calculated by fractal geometry analysis, and the analysis scale is dynamically adjusted by Lyapunov exponent to optimize the feature extraction accuracy. Finally, an enhanced feature set and a dynamic topology model are constructed to accurately identify the fault node or line location. This method effectively captures the nonlinear and multi-scale characteristics of power grid signals, improves the positioning accuracy by combining chaos analysis and graph neural networks, overcomes the limitations of traditional methods in complex power grids, provides an efficient and robust solution for power system fault diagnosis, and significantly improves the reliability and stability of power grid operation.

[0018] Of course, implementing any product of the present application does not necessarily require all the advantages described above to be achieved at the same time. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0020] Figure 1 System architecture diagram applicable to embodiments of the present application;

[0021] Figure 2 Flowchart of the power grid fault positioning method based on wavelet transform provided by the embodiments of the present application;

[0022] Figure 3 Structure block diagram of the power grid fault positioning device based on wavelet transform provided by the embodiments of the present application;

[0023] Figure 4 Schematic block diagram of an electronic device provided by the embodiments of the present application. DETAILED DESCRIPTION

[0024] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art belong to the scope of protection of the present application.

[0025] The terminology used in the description of the application herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used in the description of the application and the appended claims, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.

[0026] It should be understood that the term "and / or" as used herein merely describes associated objects, which can exist in three relationships, for example, A and / or B can represent three cases: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " herein generally represents an "or" relationship between the front and rear associated objects.

[0027] Depending on the context, the word "if" as used herein can be interpreted to mean "when" or "while" or "in response to determining" or "in response to detecting." Similarly, the phrase "if it is determined" or "if (a stated condition or event) is detected" can be interpreted to mean "when it is determined" or "in response to determining" or "when (a stated condition or event) is detected" or "in response to detecting (a stated condition or event)".

[0028] Currently, there are some technologies such as Fourier transform or simple time domain analysis, which are difficult to effectively process the non-stationary and nonlinear characteristics of power grid signals, resulting in insufficient detection sensitivity or large positioning error.

[0029] In view of this, the present application provides a new idea. In order to facilitate the understanding of the present application, first, the system architecture based on the present application is described. Figure 1 An exemplary system architecture to which embodiments of the present application can be applied is shown, as shown in Figure 1 The system architecture can include a user device and a wavelet transform-based power grid fault location device located at the server end.

[0030] The user can input voltage, current and power flow signals through the user device, and the user device sends them to the wavelet transform-based power grid fault location device at the server end. The wavelet transform-based power grid fault location device can use the method provided in the embodiments of the present application to obtain the fault location result. The server end can send the fault location result to the user terminal, and the user terminal can use the positioning result to optimize the network.

[0031] The user equipment can include, but is not limited to, a smart mobile terminal, a smart home device, a wearable device, a PC (Personal Computer), and the like. The smart mobile device can include a mobile phone, a tablet computer, a notebook computer, a PDA (Personal Digital Assistant), an Internet car, and the like. The smart home device can include a smart television, a smart refrigerator, and the like. The wearable device can include a smart watch, smart glasses, a virtual reality device, an augmented reality device, a mixed reality device, and the like.

[0032] The power grid fault positioning device based on wavelet transform can be set as an independent server, a server group, or a cloud server. The cloud server, also known as a cloud computing server or a cloud host, is a host product in a cloud computing service system, which solves the defects of large management difficulty and weak service scalability in traditional physical hosts and virtual private server (VPS) services. In addition to the architecture shown in the figure, the power grid fault positioning device based on wavelet transform can also be set in a computer terminal with strong computing power. Figure 1

[0033] It should be understood that Figure 1 The user equipment and the power grid fault positioning device based on wavelet transform in the figure are only illustrative. According to the implementation needs, there can be any number of user equipment and power grid fault positioning devices based on wavelet transform.

[0034] Figure 2 A flow chart of a power grid fault positioning method based on wavelet transform is provided for the embodiments of the present application. The method can be executed by the power grid fault positioning device based on wavelet transform in the system shown in the figure. As shown in the figure, the method can include the following steps: Figure 1 Figure 2

[0035] Step 201: Collecting voltage, current, and power flow signals of the power grid to generate a power grid dynamic signal data set containing time series features.

[0036] Step 202: Using wavelet transform based on signal local complexity optimization to perform multi-scale decomposition on the power grid dynamic signal data set to generate initial signal feature components at each scale.

[0037] Step 203: Performing fractal geometry analysis on the initial signal feature components to calculate corresponding Hurst exponents and fractal dimensions.

[0038] ​​​Step 204: Calculate the Lyapunov exponent of the initial signal characteristic component, and dynamically adjust the scale parameter of the fractal analysis according to the Lyapunov exponent, to obtain the calculation results of the optimized Hurst index and fractal dimension.

[0039] Step 205: Construct an enhanced feature set representing fault characteristics according to the calculation results of the optimized Hurst index and fractal dimension.

[0040] Step 206: Based on the enhanced feature set, construct a dynamic topology model of power grid fault, and identify the node or line location of fault occurrence according to the model, to generate fault location results.

[0041] As can be seen from the above process, the present application generates a dynamic signal data set by collecting power grid voltage, current and power flow signals, performs multi-scale decomposition using optimized wavelet transform, and extracts initial signal characteristic components. The Hurst index and fractal dimension are calculated using fractal geometric analysis, and the analysis scale is dynamically adjusted through the Lyapunov exponent to optimize the feature extraction accuracy. Finally, an enhanced feature set and a dynamic topology model are constructed to accurately identify the fault node or line location. This method effectively captures the nonlinear and multi-scale characteristics of power grid signals, improves the positioning accuracy by combining chaotic analysis and graph neural networks, overcomes the limitations of traditional methods in complex power grids, and provides an efficient and robust solution for power system fault diagnosis, significantly improving the reliability and stability of power grid operation.

[0042] First, the above step 201, i.e., "collecting voltage, current and power flow signals of the power grid to generate a power grid dynamic signal data set containing time series features", is described in detail in combination with an embodiment.

[0043] Collecting voltage, current and power flow signals of the power grid and generating a power grid dynamic signal data set containing time series features is the basic step of the present application, and its core is to obtain original data that can fully reflect the operating state of the power grid and organize it into a structured data set suitable for subsequent analysis. This process first involves collecting voltage signals, current signals and power flow signals from key nodes of the power grid, such as substations, transmission lines or distribution equipment, in real time. These signals are the core parameters of power grid operation, with voltage reflecting node potential difference, current representing charge flow, and power flow comprehensively describing the direction and size of energy transmission. Through high-frequency sampling equipment, transient changes and steady-state characteristics of the signals are captured to reflect the dynamic behavior of the power grid under normal operation or fault conditions.

[0044] During the acquisition process, high-precision sensors and data acquisition systems such as voltage transformers, current transformers, and power analyzers are used to ensure the accuracy and integrity of the signals. The acquired signals are typically recorded in time series form, with each signal point corresponding to a specific timestamp, forming continuous time series data. These data not only contain amplitude information but also imply the dynamic characteristics of power grid operation, such as periodic fluctuations, mutations, or noise interference. To ensure data quality, the acquisition stage usually includes preliminary noise filtering and signal calibration to reduce the impact of external interference or equipment errors.

[0045] Generating the power grid dynamic signal dataset is a process of structuring the acquired raw signals. Voltage, current, and power flow signals are integrated into a unified data structure, retaining time series characteristics to reflect the signal variation over time. This dataset not only contains time series information of single signals but also possibly derives features such as power factor or frequency change, further enriching data content. The structured design of the dataset facilitates subsequent wavelet transform and fractal analysis, ensuring efficient extraction of multi-scale features and fault-related information. The final generated dynamic signal dataset provides a solid foundation for subsequent multi-scale decomposition and feature analysis, and is the key starting point for precise fault location.

[0046] The following describes the step 202, i.e., "using wavelet transform based on signal local complexity optimization to perform multi-scale decomposition on the power grid dynamic signal dataset, and generate initial signal feature components at each scale," in detail with examples.

[0047] Wavelet transform, as a time-frequency analysis tool, can decompose signals into components of different frequencies and time scales, revealing the local characteristics of signals in different time windows and frequency ranges, thus adapting to the multi-scale behavior of power grid signals.

[0048] The optimization of signal local complexity is the key innovation point of this step. Local complexity can be quantified by calculating the local Shannon entropy of the signal, which reflects the information uncertainty or complexity of the signal in a specific time window. The calculation of local Shannon entropy is based on the probability distribution of signal amplitude, and the higher the entropy value, the more dramatic or disordered the signal changes in that time period. By analyzing the local Shannon entropy of the dataset, we can identify regions with high complexity in the signal, such as fault-induced mutations or high-frequency interference regions. These regions usually contain fault-related information and require more detailed decomposition scales to capture their features.

[0049] The optimization process based on local complexity is reflected in the dynamic selection of wavelet basis function and decomposition level. The wavelet basis function determines the resolution and feature extraction capability of signal decomposition. For example, the Daubechies basis function is suitable for capturing sharp changes, while the Morlet basis function is more suitable for analyzing periodic signals. According to the results of local Shannon entropy, the method dynamically selects the basis function that best suits the current signal complexity, while determining the decomposition level to balance the calculation efficiency and feature resolution. For example, in high complexity regions, more decomposition levels are used to capture fine-grained features, while in smooth regions, the number of levels is reduced to reduce the calculation cost. This adaptive optimization ensures that the wavelet transform can efficiently adapt to the local characteristics of the signal.

[0050] Specifically, first, the local Shannon entropy of the power grid dynamic signal dataset is calculated to quantify the local complexity of the signal. Assuming that the signal dataset is a time series x(t), within the time window [t i ,t i +Δt], the signal is discretized as {x1,x2,…,x N}. The local Shannon entropy H is calculated as follows:

[0051]

[0052] Where p k is the probability of signal amplitude distribution within the window, calculated based on the amplitude histogram, and K is the number of histogram bins. By sliding the window (the size Δt is usually tens to hundreds of sampling points, depending on the signal frequency), the local Shannon entropy is calculated for the entire time series to generate the entropy sequence H(t). High entropy values correspond to regions of sharp changes or complexity, such as fault-induced mutations; low entropy values indicate smooth regions.

[0053] Next, the wavelet basis function is dynamically selected based on the local Shannon entropy. Commonly used wavelet basis functions include Daubechies (dbN), Morlet, and Symlet, each of which has different advantages in capturing signal characteristics (such as sharp changes or periodicity). The selection strategy is based on entropy value threshold division:

[0054] If H(t i )>H high (low complexity threshold, e.g., 20th percentile), select the Morlet basis function, which is suitable for smooth periodic signals;

[0055] If H(t i )<H low (low complexity threshold, e.g., 20th percentile), select the Morlet basis function, which is suitable for smooth periodic signals;

[0056] For intermediate entropy values, Symlet basis functions (e.g., sym4) are chosen to balance the two properties.

[0057] In specific implementations, the mapping of entropy values to basis functions can also be achieved through predefined rules or machine learning models (e.g., decision trees). For example, the entropy value interval [H min , H max ] can be linearly mapped to the basis function parameter space.

[0058] Subsequently, the decomposition level L is determined. The decomposition level affects the feature resolution and computational cost, which needs to be dynamically adjusted according to the local complexity. Define the functional relationship between the decomposition level and the entropy value:

[0059]

[0060] where L min (min) is the minimum decomposition level (e.g., 3), L max (max) is the maximum decomposition level (e.g., 8), H min (min) and H max (max) are the minimum and maximum values of the entropy sequence, respectively. High-entropy regions require higher decomposition levels to capture fine-grained features, while low-entropy regions reduce the number of levels to improve efficiency.

[0061] When performing wavelet transform, for each time window [t i , t i + Δt], use the selected wavelet basis function ψ(t) and the decomposition level L(t i ) to perform discrete wavelet transform (DWT) on the signal x(t). The transform formula is:

[0062] c j,k j,k = ∫x(t)·ψ j,k j,k (t)dt, ψ j,k j,k (t) = 2 -j / 2 ψ(2 -j t-k) (3)

[0063] where c j,k j,k is the coefficient at the jth scale and the kth time point, and ψ j,k j,k (t) is the wavelet function. The decomposition result generates approximation coefficients (low-frequency components) and detail coefficients (high-frequency components), corresponding to the initial signal feature components at each scale.

[0064] In specific implementations, assume the signal sampling frequency is 10 kHz, and the window size Δt = 0.1 seconds (1000 sampling points). For a certain window, calculate H(t i ) = 3.5, which exceeds H high= 3.0, db4 basis function is selected, and the decomposition layer number L = 6. Through DWT, 6 sets of detail coefficients (D1-D6, capturing high to medium frequency characteristics) and 1 set of approximation coefficients (A6, capturing low frequency trend) are generated. These coefficients constitute the initial signal feature components, retaining the local characteristics of the time series.

[0065] Finally, the decomposition results of all windows are integrated to form a multi-scale initial signal feature component set. Each component corresponds to a specific scale and time range, containing high-frequency transients (such as fault mutations) and low-frequency trends (such as power drift), providing rich feature information for subsequent fractal analysis and fault location. The entire process optimizes the decomposition efficiency and feature resolution through adaptive basis function and layer number selection, ensuring effective characterization of complex power grid signals.

[0066] The above step 203, i.e., "fractal geometric analysis of initial signal feature components and calculation of corresponding Hurst index and fractal dimension", will be described in detail below in conjunction with an embodiment.

[0067] Fractal geometry analysis is a mathematical method for studying the nonlinear characteristics and self-similar structure of complex systems, widely used in signal processing, physics and engineering. It quantifies the geometric complexity of objects and the long-term correlation of time series, revealing hidden patterns and dynamic behavior. In power grid fault location, fractal geometry analysis is used to analyze the non-stationary and nonlinear characteristics of voltage, current and other signals, helping to identify abnormal patterns caused by faults. The core of fractal geometry lies in using fractal dimension and Hurst index as indicators to describe the complexity and persistence of signals from different scales, providing quantitative basis for fault feature extraction.

[0068] Fractal dimension is an indicator that describes the geometric complexity of signals or objects, reflecting their filling degree in space or time-frequency domain. Simple signals such as smooth curves have lower fractal dimension, close to an integer (such as 1 or 2), while complex signals such as fault-induced sharp fluctuations have non-integer high fractal dimension. Hurst index is used to measure the long-term correlation and persistence of time series, with a value range of 0 to 1.

[0069] Specifically, for each scale of initial signal feature component, the box-counting method is used to calculate the fractal dimension. First, the time series of the signal is divided into a grid that covers the entire signal, with grid size ε. The number of boxes required to cover the signal is denoted as N(ε), and the fractal dimension D is defined as:

[0070] D = lim(ε→0) [logN(ε) / log(1 / ε)] (4)

[0071] In practical operation, a series of decreasing grid sizes (such as ε = 1 / 2, 1 / 4, 1 / 8, etc.) are selected, and the corresponding N(ε) is calculated, and the fractal dimension is estimated by the slope of the log-log graph. Higher fractal dimension indicates that the signal has more complex geometric structure, which may correspond to the sharp fluctuation caused by fault, while lower fractal dimension reflects the more stable signal behavior.

[0072] The Hurst exponent is calculated by using the rescaled range analysis, i.e. R / S analysis. For the characteristic component time series under a certain scale, it is divided into subintervals with length τ, and the cumulative deviation range R (i.e. the difference between the maximum deviation and the minimum deviation) and the standard deviation S of each subinterval are calculated. The relationship between R / S statistic and time scale τ satisfies:

[0073] R / S ~ τ H (5) where H is the Hurst exponent, which is obtained by linear regression after logarithmic transformation. For example, if H is close to 1, the signal has strong persistence, indicating that the fault may cause long-term trend; if H is close to 0, the signal shows anti-persistence, which may correspond to random noise or transient disturbance; H ≈ 0.5 indicates random walk characteristics.

[0074] In specific implementation, the calculation of fractal dimension and Hurst exponent is performed for each scale of characteristic component respectively. For example, for high frequency component D1, a higher fractal dimension may be detected, reflecting the sharp mutation caused by fault; for low frequency component A6, a higher Hurst exponent may be obtained, indicating the long-term change trend of power flow. The calculation results form a feature vector, recording the fractal dimension and Hurst exponent of each scale, providing quantitative basis for subsequent feature optimization and fault location. This fractal analysis method can effectively capture the nonlinear characteristics of power grid signal, enhance the discrimination of fault features, and lay a foundation for accurate positioning of power grid fault.

[0075] The step 204, i.e. "calculating Lyapunov exponent of initial signal characteristic component, and dynamically adjusting scale parameter of fractal analysis according to the Lyapunov exponent to obtain calculation results of optimized Hurst exponent and fractal dimension", will be described in detail below in combination with embodiments.

[0076] This step quantifies the chaotic characteristics of the signal, dynamically optimizes the scale selection of fractal analysis, thereby improving the accuracy of Hurst exponent and fractal dimension, and providing a more reliable basis for subsequent fault feature extraction and positioning. The initial signal characteristic component is generated by optimized wavelet transform, containing multi-scale time series, covering high-frequency transient and low-frequency trend of power grid signal. The combination of Lyapunov exponent and dynamic scale adjustment can adapt to the nonlinear dynamics of the signal, ensuring that the analysis results are more consistent with the complex behavior related to fault.

[0077] Lyapunov exponent is a core indicator to measure the chaotic degree of time series, reflecting the divergence speed of adjacent trajectories in phase space over time. There are multiple methods to calculate Lyapunov exponent, suitable for different types of time series and application scenarios. These methods quantify the divergence or convergence characteristics of trajectories in phase space to assess the chaotic degree of signals. In power grid fault location, selecting the appropriate method can better adapt to the nonlinear and non-stationary characteristics of signals. A positive Lyapunov exponent indicates that the signal has chaotic characteristics and a fast divergence speed, which may correspond to the dramatic fluctuations caused by faults; zero or negative values indicate that the signal tends to be stable or periodic.

[0078] Preferably, the Wolf algorithm is used to calculate the Lyapunov exponent. For a characteristic component time series of a certain scale, its phase space trajectory is constructed, and the divergence of two trajectories with close initial distances over time is tracked. The exponent calculation formula is:

[0079] λ = (1 / (t M -t0))*Σlog(d(t i ) / d(t i-1 )) (6)

[0080] where d(t i ) represents the distance between trajectories at time t i , and t0 to t M is the observation time range. By calculating the Lyapunov exponent for each scale component, the chaotic degree can be quantified, providing a basis for subsequent scale adjustment.

[0081] The scale parameter of dynamic adjustment of fractal analysis is based on the size of the Lyapunov exponent. Fractal analysis such as box counting and R / S analysis relies on the selection of time window or grid size, and the scale parameter directly affects the accuracy of Hurst exponent and fractal dimension. If the chaotic degree of the signal is high, i.e., the Lyapunov exponent is large, a smaller analysis window should be used to capture the rapidly changing local features; if the chaotic degree is low, a larger window should be used to reflect the long-term trend. The specific adjustment strategy is: the window size w is inversely proportional to the Lyapunov exponent λ, for example:

[0082] w = w0*exp(-k*λ) (7)

[0083] where w0 is the baseline window size, and k is the adjustment coefficient, which needs to be calibrated according to the signal sampling frequency and application scenario. This adaptive adjustment ensures that the analysis scale matches the dynamic complexity of the signal, avoiding the loss of details with a large window or the introduction of noise with a small window.

[0084] The optimized Hurst exponent and fractal dimension are recalculated with the adjusted scale parameter. For each scale component, the box-counting method and R / S analysis are performed with a dynamically determined window size. For example, in the high chaos region, a smaller window can accurately capture the mutation characteristics caused by faults, improving the resolution of the fractal dimension; in the low chaos region, a larger window enhances the ability of the Hurst exponent to characterize long-range correlation. The optimized results form a more accurate feature vector, reflecting the true complexity and persistence of the signal, providing high-quality input for constructing an enhanced feature set and fault location.

[0085] The significance of this step is to introduce chaos analysis through Lyapunov exponent, overcoming the limitations of fixed scale in traditional fractal analysis. Dynamically adjusting the scale parameter makes the analysis process adaptive to the nonlinear characteristics of the signal, significantly improving the accuracy of the Hurst exponent and fractal dimension. In power grid fault location, this method can better distinguish the signal characteristics under normal operation and fault conditions, especially for non-stationary disturbances in complex power grids, laying a solid foundation for subsequent dynamic topology modeling and fault identification.

[0086] The above step 205, i.e., "constructing an enhanced feature set representing fault characteristics according to the calculation results of the optimized Hurst exponent and fractal dimension", will be described in detail below in conjunction with an embodiment.

[0087] Constructing an enhanced feature set representing fault characteristics according to the calculation results of the optimized Hurst exponent and fractal dimension is a key step in this application to convert fractal analysis results into efficient feature representation for fault location. The optimized Hurst exponent and fractal dimension are generated by the previous steps, representing the long-range correlation and geometric complexity of the power grid signal, respectively, and containing dynamic information related to faults. The construction of this feature set integrates multi-scale analysis results through nonlinear fusion and feature selection, forming a compact and high-information feature set, providing accurate input for subsequent dynamic topology modeling and fault location.

[0088] The first step in constructing the enhanced feature set is to fuse the optimized Hurst exponent and fractal dimension through nonlinear mapping. The Hurst exponent and fractal dimension describe the signal characteristics from two dimensions of time correlation and spatial complexity, respectively, and can be simply constructed into an enhanced feature set by direct concatenation.

[0089] Preferably, constructing an enhanced feature set representing fault characteristics according to the calculation results of the optimized Hurst exponent and fractal dimension comprises: performing weighted fusion of the calculation results of the optimized Hurst exponent and fractal dimension through nonlinear mapping based on a Gaussian kernel to obtain a fusion feature; and filtering a feature subset with high information content from the fusion feature using mutual information entropy analysis to form the enhanced feature set.

[0090] A nonlinear mapping method based on Gaussian kernels is employed to map both into high-dimensional feature space to enhance expressiveness. The Gaussian kernel function is defined as:

[0091] K(x,y) = exp(-||x-y||2 2 / (2σ 2 )) (8)

[0092] where x and y represent the Hurst exponent and fractal dimension values respectively, and σ is the kernel width parameter which needs to be adjusted according to signal characteristics. This mapping highlights patterns related to faults through weighted fusion, for example, high fractal dimension and high Hurst exponent combinations can indicate severe fluctuations and persistent trends caused by faults.

[0093] Subsequently, a mutual information entropy analysis is employed to select a high-information feature subset to reduce dimensionality and preserve key fault information. Mutual information quantifies the correlation between features and fault classes, with the formula:

[0094] I(X;Y) = Σp(x,y) * log(p(x,y) / (p(x)p(y))) (9)

[0095] where X is the feature vector and Y is the fault class, and p(x,y) is the joint probability distribution. By calculating the mutual information value of each fused feature with the fault class, a high-information feature subset is selected. For example, features highly correlated with fault occurrence, such as high fractal dimension in high-frequency components or high Hurst exponent in low-frequency components, are prioritized. This selection process eliminates redundant or low-correlation features, generating a compact enhanced feature set that reduces computational complexity while improving feature discrimination ability.

[0096] The following describes the step 206, i.e., "based on the enhanced feature set, constructing a dynamic topology model of the power grid fault, and identifying the node or line location where the fault occurs according to the model to generate a fault location result", in detail in combination with embodiments.

[0097] The first step in constructing the dynamic topology model is to integrate the enhanced feature set into the topology representation of the power grid, generating a weighted directed graph model. The power grid is abstracted as a graph structure, where nodes represent power grid devices such as buses, generators, or transformers, and edges represent transmission lines or connection relationships. The enhanced feature set assigns feature vectors to each node and edge, for example, node features may include the fractal dimension of the corresponding device's voltage signal, and edge features may reflect the Hurst exponent of line power flow. The weights of the graph are determined by the correlation between features or physical parameters such as line impedance, forming a weighted directed graph. This model dynamically reflects the operating state of the power grid and can capture feature changes caused by faults, such as abnormal increases in the fractal dimension of a certain node, which may indicate a local fault.

[0098] Subsequently, a graph attention network is used to analyze the feature correlation in the weighted directed graph through a multi-head attention mechanism. The Graph Attention Network (GAT) is a neural network model based on a graph structure, which is specifically used to process the node and edge relationships in graph data. It is widely used in network analysis, recommendation systems, and fault location. It dynamically assigns weights through the attention mechanism to capture the importance differences between nodes, thereby effectively modeling the interactive relationships of complex networks. In power grid fault location, the graph attention network can use the topological structure and node characteristics of the power grid to identify nodes or lines that are highly correlated with the fault and improve positioning accuracy. The graph attention network calculates the attention weight for each node and quantifies its correlation with the characteristics of neighboring nodes. The formula is:

[0099] α ij =exp(LeakyReLU(a T [Wh i ||Wh j ])) / Σexp(LeakyReLU(a T [Wh i ||Wh k ])) (10)

[0100] Among them, h i is the feature vector of node i, W is the weight matrix, and a is the attention parameter. Attention weights highlight nodes and edges that are highly correlated with fault characteristics. For example, a faulty node may receive a higher weight due to its abnormal characteristics. Based on these weights, the method can filter out the nodes or lines most likely to have faults from the complex topology of the power grid and generate accurate localization results. The specific implementation process is as follows:

[0101] Input weighted directed graph and attention weights: Get the weight α of each node i to neighbor j from the graph attention network ij , and edge weights (derived from the weights of the connecting nodes).

[0102] Calculate the total attention score of nodes and edges: For node i, calculate the total attention score S i =Σ j∈Ni α ij , where N i is the neighbor set; the opposite edge e ij , assigned to α ij or α ij The average value of the edge score S e .

[0103] Screen candidate fault locations: Set threshold T (such as 0.8), select S i >Nodes of T and S eEdges of T; combined with topological information (such as node degree or edge power flow), preferentially select high-flow area candidates; fuse multi-head attention results: average or maximum of K-head attention scores to generate comprehensive scores and

[0104] Output fault location results: rank candidate nodes and edges in descending order of scores, output the top N (e.g. N=3) as fault locations; format output: node / edge identification, fault probability, topological location (e.g. "node N3, bus A, probability 0.85").

[0105] Example: the total attention score of a certain node N5 is S_5=0.9, the connection edge L5-6 score is S_e=0.87, both are higher than the threshold 0.8, and N5 is located in a high power flow area, so it is determined as a fault node, and the output is "node N5, fault probability 0.9".

[0106] The multi-head attention mechanism further enhances the model's ability to model complex relationships by parallel computing multiple attention distributions. This method uses global information of the power grid topology to overcome the limitations of traditional local signal analysis.

[0107] Fault location is achieved by analyzing attention weights and graph model output. The graph attention network iteratively optimizes weight parameters to calculate the fault probability of each node and edge. Fault probability is based on feature correlation and topological connections, for example, a node connected to multiple high-weight edges and showing high fractal dimension in its features may be identified as a fault point. According to the pre-set probability threshold, the node or line location where the fault occurs is determined. For example, nodes or edges with a probability higher than 0.8 are marked as fault locations. Finally, the fault location results are generated, including the identification of fault nodes or lines and their probabilities, and the format is adapted to the power grid monitoring system for quick response and repair.

[0108] Preferably, the application determines the node or line location where the fault occurs according to the pre-set probability threshold, and the output of the fault location result also includes: based on the historical operation data of the power grid dynamic signal data set, a statistical model of fault probability distribution is constructed; using the statistical model, combined with the current power grid operation state, the probability threshold is dynamically adjusted; according to the adjusted probability threshold, the fault nodes or lines are reselected to generate an optimized fault location result.

[0109] Specifically, the historical data of power grid dynamic signal dataset, including time series of voltage, current and power flow, are utilized to extract the probability distribution characteristics at the time of fault occurrence. Statistical methods such as Gaussian mixture model or Bayesian estimation are adopted to analyze the distribution law of historical fault probability, forming a model describing the relationship between fault probability and power grid state. For example, the model can capture the fluctuation range of fault probability at peak load, or the probability deviation caused by noise at low load. This statistical model provides data support for dynamic threshold adjustment, ensuring that the threshold matches the actual operating environment of the power grid.

[0110] Dynamic adjustment of probability threshold is combined with the current operating state of the power grid. The operating state can be represented by real-time signal features (such as local Shannon entropy or fractal dimension) or external parameters (such as load level, weather conditions). For example, in a high load state, the fault probability may be biased high due to noise interference, and the threshold needs to be raised to reduce false positives; in a low load state, lowering the threshold can improve detection sensitivity. The adjustment formula can be defined as:

[0111] T' = T0 + α * f(S) (11)

[0112] where T0 is the initial threshold, f(S) is the adjustment factor calculated based on the statistical model and operating state S, and α is the weight coefficient. The adjusted threshold T' better reflects the dynamic characteristics of the current power grid.

[0113] According to the adjusted probability threshold, the fault nodes or lines are re-screened to generate an optimized fault location result. By comparing the fault probability output by the graph attention network with the new threshold, nodes or edges that meet the conditions are selected. For example, if the adjusted threshold is 0.85, nodes or edges with a probability higher than this value are marked as fault locations. The optimized result is output in a structured format, including fault identification, probability and topological location, which is suitable for power grid monitoring systems. This process improves the positioning accuracy, especially in complex or changing power grid environments, significantly reducing false positives and false negatives.

[0114] The above method provided by the embodiments of the present application can be applied to various application scenarios, including but not limited to: first, the method can be applied to real-time monitoring and fault positioning of large regional power grids, especially in complex topological structures, through multi-scale signal decomposition and graph attention network, the faults of power transmission lines or key nodes such as short circuit or disconnection can be accurately identified, and the method is suitable for operation and maintenance of cross-regional high-voltage power grids. Second, the method is suitable for distributed energy systems, such as microgrids containing wind power, photovoltaic and other new energy. Due to the increase in signal non-stationarity caused by the access of new energy, the method accurately captures abnormal fluctuations caused by faults through fractal analysis and Lyapunov exponent optimization, and ensures the stability of the microgrid. In addition, the method can be used for fast fault positioning of smart distribution networks, combined with dynamic topological modeling, adapting to frequent load changes and distributed device interactions in distribution networks, and reducing power outage time. The above scenarios benefit from multi-scale analysis, chaos characteristic quantification and adaptive feature extraction of the method, which can cope with the nonlinear dynamics of complex power grids and improve the robustness and practicality of fault positioning, providing strong support for the intelligent management of modern power systems.

[0115] The above describes certain embodiments of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different than the order in the embodiments and still achieve the desired result. In addition, the processes depicted in the drawings do not necessarily require the particular order shown or sequential order to achieve the desired results. In certain implementations, multi-task processing and parallel processing are possible or can be advantageous.

[0116] According to another aspect, embodiments of a wavelet transform based power grid fault positioning apparatus are provided. Figure 3 A schematic block diagram of the wavelet transform based power grid fault positioning apparatus according to one embodiment is shown. As shown, Figure 3 The apparatus 300 comprises:

[0117] A power grid data acquisition unit 301 configured to acquire voltage, current and power flow signals of a power grid, and generate a power grid dynamic signal dataset containing time series features.

[0118] A feature component generation unit 302 configured to perform multi-scale decomposition on the power grid dynamic signal dataset using wavelet transform based on signal local complexity optimization, and generate initial signal feature components at each scale.

[0119] A fractal geometry analysis unit 303 configured to perform fractal geometry analysis on the initial signal feature components, and calculate corresponding Hurst exponents and fractal dimensions.

[0120] The fractal result optimization unit 304 is configured to calculate the Lyapunov exponent of the initial signal feature component, and dynamically adjust the scale parameter of the fractal analysis according to the Lyapunov exponent, to obtain the calculation results of the optimized Hurst index and fractal dimension.

[0121] The enhanced feature set construction unit 305 is configured to construct an enhanced feature set representing fault features according to the calculation results of the optimized Hurst index and fractal dimension.

[0122] The positioning result generation unit 306 is configured to construct a dynamic topology model of the power grid fault based on the enhanced feature set, and identify the node or line position where the fault occurs according to the model, to generate a fault positioning result.

[0123] As an implementable manner, the feature component generation unit 302 can be configured to calculate the local Shannon entropy of the power grid dynamic signal data set to determine the local complexity of the signal, and dynamically select the wavelet basis function and the decomposition layer number according to the local complexity, to perform wavelet transform to generate the multi-scale initial signal feature components, when performing multi-scale decomposition on the power grid dynamic signal data set by using wavelet transform based on signal local complexity optimization.

[0124] As an implementable manner, the fractal geometry analysis unit 303 can be configured to calculate the fractal dimension of the initial signal feature component at each scale by using a box-counting-based fractal analysis method, and calculate the Hurst index by re-scaled range analysis to represent the long-range correlation of the signal, when performing fractal geometry analysis on the initial signal feature component to calculate the corresponding Hurst index and fractal dimension.

[0125] As an implementable manner, the fractal result optimization unit 304 can be configured to calculate the Lyapunov exponent of the initial signal feature component by Wolf algorithm to quantify the chaotic characteristics of the signal, and dynamically adjust the window scale of the fractal analysis according to the size of the Lyapunov exponent, so that the window size is positively correlated with the chaotic degree of the signal, when calculating the Lyapunov exponent of the initial signal feature component and dynamically adjusting the scale parameter of the fractal analysis according to the Lyapunov exponent.

[0126] As an implementable manner, the enhanced feature set construction unit 305 can be configured to perform weighted fusion on the calculation results of the optimized Hurst index and fractal dimension by using nonlinear mapping based on a Gaussian kernel to obtain fusion features, and screen a feature subset with high information quantity from the fusion features by using mutual information entropy analysis to form the enhanced feature set, when constructing the enhanced feature set representing fault features according to the calculation results of the optimized Hurst index and fractal dimension.

[0127] As an implementable manner, the positioning result generation unit 306 can be configured to, when constructing a dynamic topology model of the power grid based on the enhanced feature set and identifying the node or line location where the fault occurs according to the model: based on the enhanced feature set, construct a weighted directed graph model of the power grid, wherein the nodes represent the power grid devices and the edges represent the line connections; adopt a graph attention network to calculate the feature correlation of each node and edge in the weighted directed graph through a multi-head attention mechanism to generate dynamically allocated attention weights; calculate the fault probability of each node and line by iteratively optimizing the weight parameters of the graph attention network; determine the node or line location where the fault occurs according to a preset probability threshold, and output the fault positioning result.

[0128] As an implementable manner, the positioning result generation unit 306 can be configured to, when determining the node or line location where the fault occurs according to a preset probability threshold and outputting the fault positioning result: based on the historical operation data of the power grid dynamic signal data set, construct a statistical model of the fault probability distribution; dynamically adjust the probability threshold by using the statistical model in combination with the current power grid operation state; and according to the adjusted probability threshold, re-screen the fault nodes or lines to generate an optimized fault positioning result.

[0129] Each of the embodiments in the specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the difference from other embodiments. In particular, for the device embodiments, since they are basically similar to the method embodiments, they are described more simply, and the relevant parts can be referred to the part of the method embodiments. The above-described device embodiments are only illustrative, and the units described as separate components can be or can not be physically separated, and the components displayed as units can be or can not be physical units, that is, they can be located in one place or distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment scheme according to actual needs. Those skilled in the art can understand and implement it without creative labor.

[0130] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation portal for user to choose authorization or refusal.

[0131] In addition, the embodiment of the present application further provides a computer readable storage medium, which stores a computer program, and the program is executed by a processor to implement the steps of the method in any one of the foregoing method embodiments.

[0132] An electronic device includes one or more processors and a memory associated with the one or more processors, the memory configured to store program instructions that, when read and executed by the one or more processors, perform the steps of the method in any one of the foregoing method embodiments.

[0133] The present application also provides a computer program product comprising a computer program which, when executed by a processor, implements the steps of the method in any one of the foregoing method embodiments.

[0134] In some embodiments, the electronic device includes one or more processors and a memory associated with the one or more processors, the memory configured to store program instructions that, when read and executed by the one or more processors, perform the steps of the method in any one of the foregoing method embodiments. Figure 4 An exemplary architecture of an electronic device is shown, which can specifically include a processor 410, a video display adapter 411, a disk drive 412, an input / output interface 413, a network interface 414, and a memory 420. The processor 410, the video display adapter 411, the disk drive 412, the input / output interface 413, the network interface 414, and the memory 420 can be communicatively connected through a communication bus 430.

[0135] The processor 410 can be implemented in the form of a general-purpose CPU, a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is configured to execute related programs to implement the technical solutions provided by the present application.

[0136] The memory 420 can be implemented in the form of a ROM (Read Only Memory), a RAM (Random Access Memory), a static storage device, a dynamic storage device, etc. The memory 420 can store an operating system 421 for controlling the operation of the electronic device 400, a basic input / output system (BIOS) 422 for controlling the low-level operation of the electronic device 400. In addition, a web browser 423, a data storage management system 424, and a power grid fault positioning device based on wavelet transform 425, etc. can also be stored. The power grid fault positioning device based on wavelet transform 425 can be an application program that specifically implements the foregoing steps in the embodiments of the present application. In summary, when the technical solutions provided by the present application are implemented by software or firmware, the related program codes are stored in the memory 420 and executed by the processor 410.

[0137] The input / output interface 413 is configured to connect an input / output module to realize information input and output. The input / output module can be configured as a component in the device (not shown in the figure) or externally connected to the device to provide corresponding functions. The input device can include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output device can include a display, a speaker, a vibrator, an indicator light, etc.

[0138] The network interface 414 is configured to connect a communication module (not shown in the figure) to realize communication interaction between the device and other devices. The communication module can realize communication through a wired manner (such as a USB, a network cable, etc.) or a wireless manner (such as a mobile network, WIFI, Bluetooth, etc.).

[0139] The bus 430 includes a path for transmitting information between various components (such as the processor 410, the video display adapter 411, the disk drive 412, the input / output interface 413, the network interface 414, and the memory 420) of the device.

[0140] It should be noted that although the above device only shows the processor 410, the video display adapter 411, the disk drive 412, the input / output interface 413, the network interface 414, the memory 420, and the bus 430, in the specific implementation process, the device can also include other components necessary for normal operation. In addition, those skilled in the art can understand that the above device can also only contain the components necessary to implement the scheme of the present application, and does not have to contain all the components shown in the figure.

[0141] From the above description of the embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of software and a general hardware platform. Based on such understanding, the technical solutions of the present application can be embodied in the form of a computer program product, which can be stored in a storage medium such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments of the present application.

[0142] The technical solutions provided by the present application are described in detail above, and the principles and implementation manners of the present application are described by applying specific examples. The above description of the embodiments is only to help understand the method and its core idea of the present application; at the same time, for those skilled in the art, according to the idea of the present application, the specific implementation manner and application range will be changed. In conclusion, the content of the specification should not be understood as a limitation of the present application.

Claims

1. A power grid fault location method based on wavelet transform, characterized in that: The method comprises: Collect the voltage, current and power flow signals of the power grid and generate a power grid dynamic signal dataset containing time series characteristics; Performing multi-scale decomposition on the power grid dynamic signal data set by using wavelet transform based on signal local complexity optimization to generate initial signal characteristic components at each scale; Performing fractal geometry analysis on the characteristic components of the initial signal to calculate the corresponding Hurst exponent and fractal dimension; Calculating the Lyapunov exponent of the characteristic component of the initial signal, and dynamically adjusting the scale parameter of the fractal analysis according to the Lyapunov exponent to obtain the calculation results of the optimized Hurst exponent and fractal dimension; Constructing an enhanced feature set representing fault characteristics according to the calculation results of the optimized Hurst exponent and fractal dimension; Based on the enhanced feature set, a dynamic topology model of the power grid fault is constructed, and the node or line location where the fault occurs is identified according to the model to generate a fault location result.

2. The power grid fault location method according to claim 1, characterized in that: The power grid dynamic signal dataset is decomposed into multiple scales using wavelet transform based on signal local complexity optimization to generate initial signal feature components at each scale, including: Calculating the local Shannon entropy of the power grid dynamic signal data set to determine the local complexity of the signal; The wavelet basis function and the number of decomposition layers are dynamically selected according to the local complexity, and the wavelet transform is performed to generate multi-scale initial signal characteristic components.

3. The power grid fault location method according to claim 1, characterized in that: Performing fractal geometry analysis on the characteristic components of the initial signal to calculate the corresponding Hurst exponent and fractal dimension includes: Using a fractal analysis method based on box counting, the fractal dimensions of the characteristic components of the initial signal at each scale are calculated; The Hurst exponent was calculated by rescaled range analysis to characterize the long-range correlation of the signal.

4. The power grid fault location method according to claim 1 or 3, characterized in that: Calculating the Lyapunov exponent of the characteristic component of the initial signal and dynamically adjusting the scale parameter of the fractal analysis according to the Lyapunov exponent includes: Calculating the Lyapunov exponent of the characteristic component of the initial signal by Wolf algorithm to quantify the chaotic characteristics of the signal; According to the size of the Lyapunov exponent, the window size of the fractal analysis is dynamically adjusted so that the window size is positively correlated with the degree of chaos of the signal.

5. The power grid fault location method according to claim 1, characterized in that: The enhanced feature set for characterizing fault characteristics is constructed based on the calculation results of the optimized Hurst exponent and fractal dimension, including: By nonlinear mapping based on Gaussian kernel, weighted fusion is performed on the calculation results of the optimized Hurst exponent and fractal dimension to obtain fusion features; Mutual information entropy analysis is used to screen feature subsets with high information content in the fused features to form the enhanced feature set.

6. The power grid fault location method according to claim 1, characterized in that: Constructing a dynamic topology model of a power grid fault based on the enhanced feature set, and identifying a node or line location where the fault occurs according to the model includes: Based on the enhanced feature set, a weighted directed graph model of the power grid is constructed, wherein nodes represent power grid devices and edges represent line connections; A graph attention network is used to calculate the feature correlation of each node and edge in the weighted directed graph through a multi-head attention mechanism to generate dynamically assigned attention weights; By iteratively optimizing the weight parameters of the graph attention network, the fault probability of each node and line is calculated; according to the preset probability threshold, the node or line location where the fault occurs is determined, and the fault location result is output.

7. The power grid fault location method according to claim 6, characterized in that: The node or line location where the fault occurred is determined based on the preset probability threshold. The fault location output also includes: Building a statistical model of fault probability distribution based on historical operating data of the power grid dynamic signal data set; Using the statistical model and combining it with the current grid operation status, dynamically adjusting the probability threshold; Based on the adjusted probability threshold, the faulty nodes or lines are re-screened to generate optimized fault location results.

8. A power grid fault location device based on wavelet transform, characterized in that: The device comprises: a power grid data acquisition unit configured to acquire voltage, current, and power flow signals of the power grid and generate a power grid dynamic signal dataset containing time series characteristics; a characteristic component generating unit configured to perform multi-scale decomposition on the power grid dynamic signal data set using a wavelet transform based on signal local complexity optimization to generate initial signal characteristic components at each scale; A fractal geometry analysis unit is configured to perform fractal geometry analysis on the characteristic components of the initial signal and calculate the corresponding Hurst exponent and fractal dimension; a fractal result optimization unit configured to calculate the Lyapunov exponent of the characteristic component of the initial signal, and dynamically adjust the scale parameter of the fractal analysis according to the Lyapunov exponent to obtain the calculation results of the optimized Hurst exponent and fractal dimension; an enhanced feature set construction unit configured to construct an enhanced feature set representing fault characteristics according to the calculation results of the optimized Hurst exponent and fractal dimension; The positioning result generating unit is configured to construct a dynamic topology model of the power grid fault based on the enhanced feature set, identify the node or line location where the fault occurs according to the model, and generate a fault positioning result.

9. An electronic device, characterized in that: include: one or more processors; and a memory associated with the one or more processors, the memory being used to store program instructions, wherein the program instructions, when read and executed by the one or more processors, execute the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images